Re: linux-next: build failure on powerpc 8xx with 16k pages

[Will Deacon <will@kernel.org>]

[+Arnd since I think we spoke about this on IRC once]

On Thu, Jun 04, 2020 at 02:35:14PM +0000, Christophe Leroy wrote:
> Now I get the same issue at
> 
>    CC      mm/mincore.o
> In file included from ./include/asm-generic/bug.h:5:0,
>                  from ./arch/powerpc/include/asm/bug.h:109,
>                  from ./include/linux/bug.h:5,
>                  from ./include/linux/mmdebug.h:5,
>                  from ./include/linux/mm.h:9,
>                  from ./include/linux/pagemap.h:8,
>                  from mm/mincore.c:11:
> In function 'huge_ptep_get',
>     inlined from 'mincore_hugetlb' at mm/mincore.c:35:20:
> ./include/linux/compiler.h:392:38: error: call to '__compiletime_assert_218'
> declared with attribute error: Unsupported access size for
> {READ,WRITE}_ONCE().
>   _compiletime_assert(condition, msg, __compiletime_assert_, __COUNTER__)
>                                       ^
> ./include/linux/compiler.h:373:4: note: in definition of macro
> '__compiletime_assert'
>     prefix ## suffix();    \
>     ^
> ./include/linux/compiler.h:392:2: note: in expansion of macro
> '_compiletime_assert'
>   _compiletime_assert(condition, msg, __compiletime_assert_, __COUNTER__)
>   ^
> ./include/linux/compiler.h:405:2: note: in expansion of macro
> 'compiletime_assert'
>   compiletime_assert(__native_word(t) || sizeof(t) == sizeof(long long), \
>   ^
> ./include/linux/compiler.h:291:2: note: in expansion of macro
> 'compiletime_assert_rwonce_type'
>   compiletime_assert_rwonce_type(x);    \
>   ^
> ./include/asm-generic/hugetlb.h:125:9: note: in expansion of macro
> 'READ_ONCE'
>   return READ_ONCE(*ptep);
>          ^
> make[2]: *** [mm/mincore.o] Error 1
> 
> I guess for this one I have to implement platform specific huge_ptep_get()

Yeah, or bite the bullet and introduce proper accessors for all these
things:

	pte_read()
	pmd_read()
	pud_read()
	etc

with the default implementation pointing at READ_ONCE(), but allowing an
architecture override. It's a big job because mm/ would need repainting,
but it would have the benefit of being able to remove aggregate types from
READ_ONCE() entirely and using a special accessor just for the page-table
types.

That might also mean that we could have asm-generic versions of things
like ptep_get_and_clear() that work for architectures with hardware
update and need atomic rmw. But I'm getting ahead of myself.

Will